I just started looking at a production system with the pg_stat_statements module loaded. We started doing this after Peter Geoghegan's fantastic work on normalizing the statements made it much more useful to us, which was one motivating factor in migrating to PostgreSQL 9.2.
I was immediately struck with a wish for more stats. Specifically, I'd like stats on the spread of times for the calls - the minimum, maximum and standard deviation of the calls would be great things to have. Then we could separate queries that just generally run slowly from queries that mostly work well but have occasional spectacularly bad running times.